Solution structure of a hydrophobic analogue of the winter flounder
antifreeze protein
Edvards Liepinsh
1
, Gottfried Otting
1
, Margaret M. Harding
2
, Leanne G. Ward
2
, Joel P. Mackay
3
and A. D. J. Haymet
4
1
Karolinska Institute, Tomtebodava
¨
gen, Stockholm, Sweden;
2
School of Chemistry, University of Sydney, NSW, Australia;
3
Department of Biochemistry, University of Sydney, NSW, Australia;
4
Department of Chemistry and Institute for Molecular Design,
University of Houston, TX, USA
The solution structure of a synthetic mutant type I antifreeze
protein (AFP I) was determined in aqueous solution at
pH 7.0 using nuclear magnetic resonance (NMR) spectro-
scopy. The mutations comprised the replacement of the four
Thr residues by Val and the introduction of two additional
Lys-Glu salt bridges. The antifreeze activity of this mutant

Most studies have focused on the type I antifreeze
proteins [5], which are structurally the simplest members of
the AFPs. Fourteen type I proteins have been identified
from the right-eye flounders and sculpins [5], and these
proteins are characterized by being low M
r
, alanine rich,
a helical structures. Within this class, HPLC6 (TTTT) [11], a
37-residue sequence containing three 11-residue repeats of
ThrX
2
AsxX
7
is by far the most extensively studied protein
and is the only type I AFP for which a solid state structure
has been reported. Single X-ray diffraction [8,12] showed
that in the solid state this protein is completely a helical in
conformation with the exception of the last unit, which
adopts a 3
10
-helix conformation. The protein has also been
studied by NMR spectroscopy [13] but due to the high
number of alanine residues in the sequence, which led to
significant spectral overlap, full resonance assignments were
not possible. These studies confirmed the global helical
conformation of the peptide and allowed the rotamer
conformations of a number of residues to be determined,
but clear evidence for the presence of helix-stabilizing
interactions arising from the capping motifs observed in the
crystal structure was not obtained. More recently, meas-

(Received 28 September 2001, revised 21 December 2001, accepted
7 January 2002)
Eur. J. Biochem. 269, 1259–1266 (2002) Ó FEBS 2002
hydrogen bonds involving the hydroxyl groups of the four
Thr residues is not the primary reason for the interaction of
TTTT with the ice/water interfacial region.
We have recently designed and synthesized analogues of
TTTT in which the relative size, hydrophobicity and
hydrogen bonding characteristics of the side chains at
positions 2, 13, 24 and 37 were systematically varied [21,22].
Four additional charged residues K7, E11, K29 and E33
(italicized in the VVVV2KE sequence shown in Table 1)
were incorporated into the sequence to improve solubility
and minimize aggregation. The valine-substituted analogue
VVVV2KE showed similar behaviour to TTTT at low
concentrations [22] and showed conclusively that models for
the mechanism of ice growth inhibition that are dominated
by hydrogen bonding involving the Thr hydroxyls are
incorrect.
This paper reports determination of the solution structure
of VVVV2KE. The additional charged residues in this
sequence provided chemical shift dispersion in the alanine-
rich segments compared with TTTT and thus allowed the
first solution structure of a type I protein to be determined.
Such experimental solution data are important in modelling
the interaction of these peptides with the ice/water interface,
in order to provide a mechanism for the selective interaction
of the peptide with the [2 0

22 1] ice plane, and hence to

and 15 °C to support the resonance assignment and check
for conformational differences. The in-phase lineshape of
NOESY cross peaks was used to determine J
HN,Ha
coupling
constants [26]. The COSY and TOCSY cross-peaks were
visually inspected to determine the relative magnitude of the
J
Ha,Hb
couplings of C
b
H
2
methylene groups. The ROESY
spectrum at 15 °C (mixing time 50 ms) was used to identify
spin-diffusion cross-peaks in the NOESY spectrum recor-
ded with an 80-ms mixing time. The program
XEASY
was
used for resonance assignments and peak integration [27].
DYANA
[28] and
OPAL
[29] were used for the structure
calculations and energy minimization, respectively. Stand-
ard parameters were used for both programs. The energy
minimization was performed in a water shell of 6 A
˚
.
Hydrogen bonds were identified by O ÆÆÆH distances

sector cells, and data were collected at 0 °CinanAn-60ti
rotor (45 000 and 54 000 r.p.m.). Data were acquired as
absorbance vs. radius scans (at 240 and 360 nm) at 0.001-cm
intervals and as the sum of 10 scans. Data were collected at
3-h intervals and compared to determine when the samples
had reached chemical and sedimentation equilibrium. After
subtraction of the 360-nm scans, the data from all speeds
and loading concentrations were fitted simultaneously to a
number of models using the program
NONLIN
[31]; the
quality of each fit was determined by inspection of residual
plots and v
2
values. Visualization of the plots of apparent
molecular mass vs. concentration and W vs. concentration
was carried out using the program
OMMENU
[32].
RESULTS
Analytical ultracentrifugation
Figure 1 shows the results of AU experiments on
VVVV2KE at three concentrations, including the fitted
curves obtained using an ideal single species model. The
combined residuals of the fit are presented in the bottom
panel of Fig. 1. The derived molecular mass for the peptide
shows that in the concentration range less than 1 m
M
,and
under the conditions used for these measurements, the

observed for example for the homologous repeats Glu11–
Ala14 and Glu22–Ala25, upper distance limit restraints
were derived using the assumption that corresponding
NOEs from the different segments contributed equally to
the overlapping cross peak intensity. Similarly, the same
dihedral angle restraints were used for homologous repeats,
when the corresponding COSY cross peaks overlapped, but
their assignment was otherwise unambiguous. The use of
identical restraints for homologous, spectrally unresolved
peptide segments was motivated by the observation of
similar NOEs and coupling constants, when cross peaks
between homologous repeats could be resolved.
NOEs with the terminal amino-acid residues were very
weak, presumably due to increased mobility. Therefore, the
set of upper distance restraints of residues 2 and 37 was
supplemented by restraints obtained from the ROESY
spectrum recorded at 15 °C and a much higher sample
concentration. Furthermore, a hydrogen bond between the
carboxyl group of Asp1 and the amide proton of Ser4 was
indicated by the observation of a large high-field shift of this
H
N
resonance when the pH was lowered to pH 2 (data not
shown) [33,34]. This hydrogen bond seems to be highly
populated at neutral pH, where the H
N
resonance of Ser4 is
the most low-field shifted amide (Fig. 2).
Solution structure of VVVV2KE
The solution structure of VVVV2KE, represented by the

these locations of the a helix. While local flexibility would
necessarily affect the amplitude and precise direction of the
helical bend calculated from NOE data, a bend of the helix
seems to be a genuine feature of VVVV2KE. The hydrogen
bond between the side chain of Asp1 and the backbone
amide of Ser4 results in the presence of an N-cap (Fig. 3B).
When this hydrogen bond was removed from the list of
restraints, it was found only in a minority of the conformers.
The presence of this hydrogen bond was, however, strongly
supported by the chemical shift changes observed in the pH
titration and it was consequently included as a restraint. The
chemical shift of Ser4 H
N
showed the largest temperature
coefficient of all amide protons (0.017 p.p.m. per °C
between 5 and )5 °C), suggesting that this hydrogen bond
is particularly short or is readily broken at higher temper-
atures. In contrast, the experimental evidence for the
presence of a well-defined C-cap, as in the crystal structure
of TTTT [8], was less clear. Any NOEs involving the
terminal residue Arg37 were weak, probably due to
increased mobility, and the temperature coefficients of the
chemical shifts of the C-terminal NH
2
group of Arg37 were
too large to suggest any involvement in a stable hydrogen
bond. Yet, the temperature coefficients of the two NH
2
protons were significantly different and smaller for the high-
field shifted proton, which in the crystal structure of TTTT

and TTTT
The crystal structure of TTTT contains two conformers in
the unit cell that differ widely in their helical bend (Fig. 4)
[8]. In the following, we refer to the more bent conformer as
the Ôb-conformerÕ, and the less bent conformer as the
Ôs-conformerÕ. Interestingly, the b- and s-conformers are
bent in opposite directions. The overall bend observed in the
NMR structure of VVVV2KE is in the same direction as in
the b-conformer, placing residues 2, 13, 24 and 37, that are
putatively involved in ice-binding, on the concave surface.
The two conformers of TTTT also differ by the side chain
v
1
rotamers of several residues, namely Asp1, Leu12, Lys18,
Leu23 and Thr35. Both conformers display the backbone
hydrogen bonds expected for an a helix spanning all
residues, and include elaborate terminal cap structures. As
with the NMR structure of VVVV2KE, the N-terminal cap
structure of TTTT includes a hydrogen bond between the
side chain carboxyl group of Asp1 and the backbone amide
of Ser4. The C-terminal cap structure, however, makes use
of the Arg37 side chain to form a hydrogen bond to the
backbone carbonyl oxygen of Ala33 [12]. No evidence of
this could be obtained in solution. Interestingly, the
Table 2.
1
H-NMR chemical shifts of VVVV2KE at 10 °C, pH 7.0. The chemical shifts were referenced to the water signal at 4.994 p.p.m. The
estimated error is ± 0.01 p.p.m. The chemical shift values of stereospecifically assigned protons are in italics, where the first number is the shift of
the proton with the lower branch number, e.g. the b
1

H
2
1.70; C
e
H
2
2.96
Ala8 8.04 4.19 1.50
Ala9 8.15 4.17 1.48
Ala10 8.00 4.18 1.52
Glu11 8.30 4.07 2.16, 2.03 C
c
H
2
2.27, 2.50
Leu12 7.83 4.26 1.70, 1.84 H
c
1.65; C
d1
H
3
0.90, C
d2
H
3
0.93
Val13 7.75 3.70 2.14 C
c1
H
3

c
H
2
2.26, 2.51
Leu23 7.82 4.26 1.70, 1.84 H
c
1.65; C
d1
H
3
0.89, C
d2
H
3
0.93
Val24 7.76 3.70 2.14 C
c1
H
3
0.96, C
c2
H
3
1.09
Ala25 7.91 4.22 1.48
Ala26 8.43 4.18 1.53
Asn27 8.67 4.55 2.79, 2.94 H
d21
7.69, H
d22

H
3
1.03
Ala36 7.93 4.18 1.43
Arg37 7.90 4.18 1.84, 1.87 C
c
H
2
1.65, 1.74; C
d
H
2
3.17,3.20; H
e
7.24;
NH
2
7.24, 7.27
1262 E. Liepinsh et al. (Eur. J. Biochem. 269) Ó FEBS 2002
chemical shift difference between the
1
H-NMR resonances
of the C-terminal NH
2
group increased by about 0.1 p.p.m.
as the temperature was lowered to )2 °C (data not shown),
suggesting that a hydrogen bond between Arg37 NH
2
and
the carbonyl oxygen of residue 35 may be significantly

The molecular mechanism whereby TTTT and other type I
proteins are able to inhibit ice growth via accumulation at
the specific [2 0

22 1] plane remains a continued subject of
discussion in the literature [5,6,24,35–37]. The first molecu-
lar dynamics simulation of a complete ice/TTTT/water
system, that does not restrict ice lattice positions, and
includes long-range electrostatic interactions, has been
reported very recently [24]. This study has allowed a
comparison of the hydrogen bonding between the protein in
water and the protein in the ice/water interfacial region.
A
B
Fig. 2. Selected spectral regions from the NOESY spectrum of
VVVV2KE in 90% H
2
O/10% D
2
Oat10°C, pH 7.0. The spectrum
was recorded at a
1
H-NMR frequency of 800 MHz, using a mixing
time of 80 ms. Cross peaks are labelled with the residue numbers of the
amino acids involved. The first/second number refers to the residue in
the d
1
/d
2
frequency dimension, respectively. (A) Cross peaks between

(right panel) in the NMR structure of VVVV2KE. The backbone
atoms of the first five and last six residues, respectively, were super-
imposed for minimum r.m.s.d. Only bonds with backbone atoms and
backbone carbonyl atoms are displayed, except for the side chain of
Asp1. The N- and C-terminal ends are identified and hydrogen bonds
drawnwithdottedlines.TheN-caphydrogenbondbetweenthe
carboxyl group of Asp1 and the backbone amide of Ser4 is identified in
bold.
Ó FEBS 2002 NMR structure of type I antifreeze protein (Eur. J. Biochem. 269) 1263
In parallel, recent experimental data on mutants that
incorporate systematic changes in both hydrophobicity
and hydrogen bonding characteristics have assisted in
defining the characteristics of the residues that are crucial
for activity and has led to new proposals for the Ôice-bindingÕ
face of the protein [22,36,37]. Further molecular dynamics
studies are required to explain these new experimental
results with mutants and to explain why TTTT recognizes
and accumulates at the {2 0

22 1} planes of ice 1h the usual
form of hexagonal ice at 1 atm.
The starting point for almost all simulations to date
[16,18,24,35,38,39] has been the X-ray coordinates of TTTT
[12]. The protein is assumed to adopt a very similar
geometry in solution, and NMR studies on TTTT are
consistent with an a helical geometry [13]. Simulations of
VVVV2KE with the ice/water interface should provide
significant insight into the mechanism of ice-growth inhibi-
tion, as this is the first example of an active mutant that
lacks hydrogen bonding side chains at positions 2, 12, 24

shifts and their temperature coefficients suggest that the
VVVV2KE structure is bent in the same direction as the
more strongly bent b-conformer in the TTTT crystal
structure [8]. Superficially the bend seems to be strongest
near Lys18 in both VVVV2KE and TTTT. As VVVV2KE
contains two additional Lys-Glu salt bridges, bends near the
additional lysines would also be expected. Indeed, the
backbone hydrogen bond between Lys29 and Ala25 is
formed in only half of the 20 NMR conformers of
VVVV2KE, but the resulting bend does not affect the
overall structure as much as that near Lys18, because Lys29
is close to the C-terminal end of the peptide. The same is true
for Lys7 near the N-terminal end, although this residue
forms correct backbone hydrogen bonds to Ala3 in all but
four of the NMR conformers.
While the overall bend in the b-conformer of TTTT is
accompanied by changes in the v
1
angles of several residues,
the side-chain conformations in the NMR structure of
VVVV2KE are more similar to those of the s-conformer.
These data can be reconciled by a model where helix
bending is facile, proceeding independently of side chain
conformations. AFP I peptides in solution would thus be
involved in an equilibrium between straight and bent helices
and, independently, equilibria between different side chain
conformations. Notably, the conformational spread among
the NMR conformers is merely a measure of the precision
with which the restraints define the structure, i.e. the
conformers are not meant to sample the entire conforma-

Maximum dihedral-angle restraint violations (°) 1.6 ± 0.3
Rmsd to the mean for N, C
a
and C¢ (A
˚
)
b
0.53 ± 0.15
Rmsd to the mean for all heavy atoms (A
˚
)
b
0.88 ± 0.16
Ramachandran plot appearance
c
Most favoured regions (%) 99.7
Additionally allowed regions (%) 0.3
Generously allowed and disallowed regions (%) 0.0
a
33
3
J(H
N,Ha
), 24
3
J(
Ha,Hb
).
b
For all residues.

dipolar couplings [40].
CONCLUSIONS
The solution structure of VVVV2KE provides an improved
basis for simulations of possible ice-binding modes.
Furthermore, the availability of sequence-specific resonance
assignments paves the way for a site-specific study of water–
peptide interactions at subzero temperatures by the use of
intermolecular water-peptide NOEs [41]. Such a study,
which can be performed in solution, seems particularly
interesting in view of the fact that the interaction of water
with the putative ice-binding surface of TTTT in the single
crystal is severely hindered by intermolecular contacts
between different peptide molecules in the crystal lattice [8].
Note added in proof: the amide chemical shift changes
and helix bend in VVVV2KE are supported by a recent
publication [Cicrpicki, T. & Otlewski, J. (2001) Amide
proton temperature coefficients as hydrogen bond indica-
tors in proteins. J. Biomol. NMR 21, 249–261], which has
shown that the temperature coefficients of the amide
chemical shifts are particularly large on the concave face
of curved helices.
ACKNOWLEDGEMENTS
ThisresearchwassupportedinpartbyanAustralianResearchCouncil
Grant (A. D. J. H. and M. M. H.), University of Sydney Sesqui
Research and Development Grant (M. M. H), Welch Grant
(A. D. J. H.) and the Swedish Research Council (E. L. and G. O.).
REFERENCES
1. DeVries, A.L. (1983) Antifreeze peptides and glycopeptides in
cold-water fishes. Ann. Rev. Physiol. 45, 245–260.
2. Davies,P.L.&Hew,C.L.(1990)Biochemistryoffishantifreeze

family. Biochemistry 37, 4712–4721.
11. Fourney, R.M., Joshi, S.B., Kao, M.H. & Hew, C.L. (1984)
Heterogeneity of antifreeze polypeptides from the Newfoundland
winter flounder, Pseudopleuronectes americanus. Can. J. Zool. 62,
28–33.
12. Sicheri, F. & Yang, D.S.C. (1995) Ice-binding structure and
mechanism of an antifreeze protein from winter flounder. Nature
375, 427–431.
13. Gronwald, W., Chao, H., Reddy, D.V., Davies, P.L., Sykes, B.D.
&So
¨
nnichsen, F.D. (1996) NMR characterization of side
chain flexibility and backbone structure in the type I antifreeze
protein at near freezing temperatures. Biochemistry 35, 16698–
16704.
14. Graether, S.P., Slupsky, C.M., Davies, P.L. & Sykes, B.D. (2001)
Structure of type I antifreeze protein and mutants in supercooled
water. Biophys. J. 81, 1677–1683.
15. Knight, C.A., Cheng, C H.C. & DeVries, A.L. (1991) Adsorption
of a-helical antifreeze peptides on specific ice crystal surface
planes. Biophys. J. 59, 409–418.
16. Chou, K.C. (1992) Energy-optimised structure of antifreeze pro-
tein and its binding mechanism. J. Mol. Biol. 223, 509–517.
17.Wen,D.&Laursen,R.A.(1992)Amodelforbindingofan
antifreeze polypeptide to ice. Biophys. J. 63, 1659–1662.
Table 4. Side chain rotamers in the NMR structure of VVVV2KE and
the X-ray structure of TTTT. Only those residues are listed that are
none-alanine in both sequences. The v
1
-angles listed are those of the

conformer and the second to the bent conformer.
b
Thr in TTTT.
c
Predominant rotamer. All three rotamers are populated.
Ó FEBS 2002 NMR structure of type I antifreeze protein (Eur. J. Biochem. 269) 1265
18. Cheng, A.L. & Merz, K.M. (1997) Ice-binding mechanism of
winter flounder antifreeze proteins. Biophys. J. 73, 2851–2873.
19. Chao, H., Houston, M.E., Hodges, R.S., Kay, C.M., Sykes, B.D.,
Loewen, M.C., Davies, P.L. & So
¨
nnichsen, F.D. (1997)
A diminished role for hydrogen bonds in antifreeze protein
binding to ice. Biochemistry 36, 14652–14660.
20. Haymet, A.D.J., Ward, L.G., Harding, M.M. & Knight, C.A.
(1998) Valine substituted winter flounder antifreeze – preservation
of ice growth hysteresis. FEBS Lett. 430, 301–306.
21. Haymet, A.D.J., Ward, L.G. & Harding, M.M. (2001) Hydro-
phobic analogues of the winter flounder antifreeze protein. FEBS
Lett. 491, 285–288.
22. Haymet, A.D.J., Ward, L.G. & Harding, M.M. (1999) Winter
flounder ÔantifreezeÕ proteins: synthesis and ice growth inhibition
of analogues that probe the relative importance of hydrophobic
and hydrogen-bonding interactions. J. Am. Chem. Soc. 121,
941–948.
23. Zhang, W. & Laursen, R.A. (1998) Structure–function relation-
ships in a type I antifreeze polypeptide – the role of threonine
methyl and hydroxyl groups in antifreeze activity. J. Biol. Chem.
273, 34806–34812.
24. Dalal, P., Knickelbein, J., Haymet, A.D.J., So

program DYANA. J. Mol. Biol. 273, 283–298.
29. Luginbu
¨
hl, P., Gu
¨
ntert, P., Billeter, M. & Wu
¨
thrich, K. (1996) The
new program OPAL for molecular dynamics simulations and
energy refinements of biological macromolecules. J. Biomol. NMR
8, 136–146.
30.Koradi,R.,Billeter,M.&Wu
¨
thrich,K.(1996)MOLMOL:a
program for display and analysis of macromolecular structures.
J. Mol. Graphics 14, 29–32.
31. Johnson, M.L., Correia, J.J., Yphantis, D.A. & Halvorson, H.R.
(1981) Analysis of data from the analytical ultracentrifuge by
nonlinear least-squares techniques. Biophys. J. 36, 575–588.
32. Ralston, G. (1994) OMMENU Programme. University of Sydney,
Australia.
33. Bundi, A. & Wu
¨
thrich, K. (1979) Proton NMR parameters of the
common amino acid residues measured in aqueous solutions of
the linear tetrapeptides H-Gly-Gly-X-L-Ala-OH. Biopolymers 18,
285–297.
34. Bundi, A. & Wu
¨
thrich, K. (1979) Use of amide proton NMR

ckert, M., Levitt, M.H. & Moshref, A. (2000)
NMR experiments for the sign determination of homonuclear
scalar and residual dipolar couplings. J. Biomol. NMR 16,
343–346.
41. Otting, G. (1997) NMR studies of water bound to biological
molecules. Prog. NMR Spectrosc. 31, 259–285.
42. Laskowski, R.A.,Rullmannn, J.A., MacArthur, M.W., Kaptein, R.
& Thornton, J.M. (1996) AQUA and PROCHECK-NMR:
programs for checking the quality of protein structures solved by
NMR. J. Biomol. NMR 8, 477–486.
1266 E. Liepinsh et al. (Eur. J. Biochem. 269) Ó FEBS 2002